In-Lane Obstacle Avoidance Using Parallel-Track Cost Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current self-driving technologies face challenges in safely and efficiently avoiding obstacles that slightly intrude into a vehicle's lane, as traditional methods either require lane changes, which introduce uncontrollable factors, or slow traffic efficiency by stopping, without adequately addressing safety and smoothness.
Innovation Solution
An obstacle avoidance method that processes vehicle and obstacle information to create a potential energy grid map, allowing the vehicle to select a target track within its lane by calculating cost values for parallel-line tracks and performing time-domain filtering to decouple the target track from the vehicle's current position, ensuring safety and smoothness by reducing sensor precision dependence and predicting potential collisions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If obstacle avoidance through lane change is performed, then collision avoidance is achieved, but uncontrollable factors from adjacent lane vehicles increase and traffic efficiency decreases
Solution Approach 1:
The patent segments the obstacle avoidance problem into two distinct approaches: avoidance within the current lane and avoidance through lane change. By segmenting the solution space, the system can selectively apply the most appropriate method based on specific scenario conditions, thereby avoiding the drawbacks of indiscriminately using lane changes for all obstacle situations.
Solution Approach 2:
The patent dynamically selects between different avoidance strategies based on real-time scenario assessment. The system evaluates obstacle characteristics, lane conditions, and traffic flow to dynamically determine whether to perform avoidance within the current lane or initiate a lane change, making the avoidance approach adaptive rather than fixed.
2Reliability
If stopping or low-speed following is performed when obstacle slightly intrudes lane, then collision is avoided, but traffic efficiency of the lane is affected
Solution Approach 1:
The patent changes the parameter of avoidance location from fixed (either stop or lane change) to variable (within current lane or adjacent lane). By adjusting this parameter based on obstacle intrusion degree and lane conditions, the system optimizes the balance between safety and traffic efficiency, allowing continued forward motion within the lane when possible.
Solution Approach 2:
The patent performs preliminary assessment of the obstacle situation before determining the avoidance action. By evaluating the obstacle's intrusion degree, position, and characteristics in advance, the system can proactively select the most efficient avoidance method, preventing unnecessary stopping or lane changes that would reduce traffic flow.
3Speed
If target track is directly calculated from current vehicle position, then real-time responsiveness is improved, but system oscillation occurs due to control delay and execution mechanism response time
Solution Approach 1:
The patent performs preliminary calculation of the target track based on the vehicle's position at the previous moment (t-1) rather than the current moment. This preliminary action anticipates the vehicle's future position, compensating for control delays and execution response times, thereby preventing system oscillation while maintaining real-time responsiveness.
Solution Approach 2:
The patent cushions against potential system oscillation by decoupling the target track calculation from the current vehicle position. By using historical position data and predicting future position, the system creates a buffer that absorbs the effects of control delays and execution mechanism response times, preventing instability.
Data Source
AI summary
Methods, apparatus, and computer programs stored on a computer-readable storage medium for obstacle avoidance are provided. One example method includes obtaining vehicle information, obstacle information, and a drivable area of a vehicle at a moment t, and processing the drivable area based on the vehicle information and the obstacle information to obtain a potential energy grid map. The method can include obtaining information about a center line of a lane in which the vehicle is currently located, and obtaining a parallel-line track cluster based on the information about the lane center line. The method can include calculating a cost value of each parallel-line track, and performing time-domain filtering on a parallel-line track with a minimum cost value in the parallel-line track cluster to obtain a target track at the moment t specifying a parallel line track that is used by the vehicle to avoid an obstacle.